audio: webrtc_aec: dynamically resolve ref source and drop excess frames

Re-resolve mic and playback-reference sources in webrtc_aec_process()
and webrtc_aec_is_ready_to_process() because IPC4 playback pipelines may
bind or unbind dynamically while capture is active, and drop excess
reference frames so playback backpressure does not stall capture or overflow
the reference ring buffer.

Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

Author: lrgirdwo

src/audio/webrtc_aec/webrtc_aec.c

@@ -168,6 +168,67 @@ static void s16_to_sink(struct sof_sink *dst, int16_t src[][WEBRTC_AEC_FIFO_FRAM
 	sink_commit_buffer(dst, nbytes);
 }
 
+static void src_drop_frames(struct sof_source *src, int frames)
+{
+	size_t frame_bytes = source_get_frame_bytes(src);
+	const void *data_ptr, *buf_start;
+	size_t buf_size, nbytes;
+
+	if (frames <= 0 || !frame_bytes)
+		return;
+
+	nbytes = (size_t)frames * frame_bytes;
+	if (!source_get_data(src, nbytes, &data_ptr, &buf_start, &buf_size))
+		source_release_data(src, nbytes);
+}
+
+static void webrtc_aec_resolve_sources(struct sof_source **sources, int num_of_sources,
+				       struct sof_sink **sinks, int num_of_sinks,
+				       int *mic_idx, int *ref_idx)
+{
+	*mic_idx = 0;
+	*ref_idx = -1;
+
+	if (num_of_sources <= 1)
+		return;
+
+#ifdef CONFIG_IPC_MAJOR_4
+	{
+		uint32_t q0 = IPC4_SINK_QUEUE_ID(source_get_id(sources[0]));
+		uint32_t q1 = IPC4_SINK_QUEUE_ID(source_get_id(sources[1]));
+
+		if (q0 != q1) {
+			if (q0 == 0) {
+				*mic_idx = 0;
+				*ref_idx = 1;
+			} else {
+				*mic_idx = 1;
+				*ref_idx = 0;
+			}
+			return;
+		}
+	}
+#endif
+
+	if (num_of_sinks > 0) {
+		if (source_get_pipeline_id(sources[0]) == sink_get_pipeline_id(sinks[0]) &&
+		    source_get_pipeline_id(sources[1]) != sink_get_pipeline_id(sinks[0])) {
+			*mic_idx = 0;
+			*ref_idx = 1;
+			return;
+		}
+		if (source_get_pipeline_id(sources[1]) == sink_get_pipeline_id(sinks[0]) &&
+		    source_get_pipeline_id(sources[0]) != sink_get_pipeline_id(sinks[0])) {
+			*mic_idx = 1;
+			*ref_idx = 0;
+			return;
+		}
+	}
+
+	*mic_idx = 0;
+	*ref_idx = 1;
+}
+
 /* ----------------------------------------------------------------
  * module_interface operations
  * ---------------------------------------------------------------- */
@@ -226,21 +287,8 @@ __cold static int webrtc_aec_prepare(struct processing_module *mod,
 		return -EINVAL;
 	}
 
-	if (num_of_sources == 1) {
-		cd->mic_src = 0;
-		cd->ref_src = -1;
-	} else if (source_get_pipeline_id(sources[0]) == sink_get_pipeline_id(sinks[0]) &&
-	    source_get_pipeline_id(sources[1]) != sink_get_pipeline_id(sinks[0])) {
-		cd->mic_src = 0;
-		cd->ref_src = 1;
-	} else if (source_get_pipeline_id(sources[1]) == sink_get_pipeline_id(sinks[0]) &&
-		   source_get_pipeline_id(sources[0]) != sink_get_pipeline_id(sinks[0])) {
-		cd->mic_src = 1;
-		cd->ref_src = 0;
-	} else {
-		cd->mic_src = 0;
-		cd->ref_src = 1;
-	}
+	webrtc_aec_resolve_sources(sources, num_of_sources, sinks, num_of_sinks,
+				   &cd->mic_src, &cd->ref_src);
 
 	mic_fmt  = source_get_frm_fmt(sources[cd->mic_src]);
 	mic_rate = source_get_rate(sources[cd->mic_src]);
@@ -344,29 +392,52 @@ static int webrtc_aec_process(struct processing_module *mod,
 			      struct sof_sink **sinks, int num_of_sinks)
 {
 	struct webrtc_aec_comp_data *cd = module_get_private_data(mod);
-	struct sof_source *mic = sources[cd->mic_src];
-	struct sof_source *ref = cd->ref_src >= 0 ? sources[cd->ref_src] : NULL;
-	struct sof_sink   *out = sinks[0];
-	int fmic = (int)source_get_data_frames_available(mic);
-	int fref = ref ? (int)source_get_data_frames_available(ref) : 0;
-	int n, drain, c;
-
-	if (fmic <= 0)
+	struct sof_source *mic;
+	struct sof_source *ref;
+	struct sof_sink   *out;
+	int mic_idx, ref_idx;
+	int fmic, fref, n, drain, c;
+	bool ref_s32;
+
+	if (num_of_sources < 1 || num_of_sinks < 1)
 		return 0;
 
-	/* Cap to available space in input accumulation buffers */
+	webrtc_aec_resolve_sources(sources, num_of_sources, sinks, num_of_sinks,
+				   &mic_idx, &ref_idx);
+	cd->mic_src = mic_idx;
+	cd->ref_src = ref_idx;
+
+	mic = sources[mic_idx];
+	ref = ref_idx >= 0 ? sources[ref_idx] : NULL;
+	out = sinks[0];
+	fmic = (int)source_get_data_frames_available(mic);
+	fref = ref ? (int)source_get_data_frames_available(ref) : 0;
+
+	/* Cap to available space in input accumulation and output sink buffers */
 	n = MIN(fmic, WEBRTC_AEC_FIFO_FRAMES - cd->buffered_in_frames);
-	if (n <= 0)
+	n = MIN(n, (int)sink_get_free_frames(out));
+	if (n <= 0) {
+		if (ref && fref > 0) {
+			size_t fb = source_get_frame_bytes(ref);
+			int keep = fb ? (int)(source_get_min_available(ref) / fb) : 0;
+
+			if (fref > keep)
+				src_drop_frames(ref, fref - keep);
+		}
 		return 0;
+	}
 
 	/* Convert mic input to per-channel S16 */
 	src_to_s16(mic, n, cd->mic_buf, cd->buffered_in_frames, cd->channels, cd->is_s32);
 
 	/* Convert ref input to per-channel S16; if starved/silent, fill with silence */
+	ref_s32 = ref ? (source_get_frm_fmt(ref) == SOF_IPC_FRAME_S32_LE) : cd->is_s32;
 	if (fref >= n) {
-		src_to_s16(ref, n, cd->ref_buf, cd->buffered_in_frames, cd->channels, cd->is_s32);
+		if (fref > n)
+			src_drop_frames(ref, fref - n);
+		src_to_s16(ref, n, cd->ref_buf, cd->buffered_in_frames, cd->channels, ref_s32);
 	} else if (fref > 0) {
-		src_to_s16(ref, fref, cd->ref_buf, cd->buffered_in_frames, cd->channels, cd->is_s32);
+		src_to_s16(ref, fref, cd->ref_buf, cd->buffered_in_frames, cd->channels, ref_s32);
 		for (c = 0; c < cd->channels; c++)
 			memset(&cd->ref_buf[c][cd->buffered_in_frames + fref], 0,
 			       (size_t)(n - fref) * sizeof(int16_t));
@@ -649,17 +720,27 @@ static bool webrtc_aec_is_ready_to_process(struct processing_module *mod,
 					   int num_of_sinks)
 {
 	struct webrtc_aec_comp_data *cd = module_get_private_data(mod);
-	struct sof_source *mic;
+	struct sof_source *mic, *ref;
 	struct sof_sink *out;
+	int mic_idx, ref_idx;
 
-	if (!cd || !cd->configured || num_of_sources <= cd->mic_src || num_of_sinks < 1)
+	if (!cd || !cd->configured || num_of_sources < 1 || num_of_sinks < 1)
 		return false;
 
-	mic = sources[cd->mic_src];
+	webrtc_aec_resolve_sources(sources, num_of_sources, sinks, num_of_sinks,
+				   &mic_idx, &ref_idx);
+	mic = sources[mic_idx];
+	ref = ref_idx >= 0 ? sources[ref_idx] : NULL;
 	out = sinks[0];
 
-	return source_get_data_available(mic) >= source_get_min_available(mic) &&
-	       sink_get_free_size(out) >= sink_get_min_free_space(out);
+	if (source_get_data_available(mic) >= source_get_min_available(mic) &&
+	    sink_get_free_size(out) >= sink_get_min_free_space(out))
+		return true;
+
+	if (ref && source_get_data_available(ref) >= 2 * source_get_min_available(ref))
+		return true;
+
+	return false;
 }
 
 static const struct module_interface webrtc_aec_interface = {